[PATCH v16 9/9] docs: add managed_irq_strict flag to isolcpus

[Aaron Tomlin]

From: Daniel Wagner <wagi@xxxxxxxxxx>

The managed_irq_strict flag informs multiqueue device drivers where to
place hardware queues and managed interrupt vectors. Document this new
flag in the isolcpus command-line argument description, describe its
relationship and precedence with respect to the legacy managed_irq flag,
and document HK_TYPE_MANAGED_IRQ_STRICT in the core housekeeping guide.

Signed-off-by: Daniel Wagner <wagi@xxxxxxxxxx>
Reviewed-by: Hannes Reinecke <hare@xxxxxxx>
Co-developed-by: Aaron Tomlin <atomlin@xxxxxxxxxxx>
Signed-off-by: Aaron Tomlin <atomlin@xxxxxxxxxxx>
---
 .../admin-guide/kernel-parameters.txt         | 37 ++++++++++++++++++-
 Documentation/core-api/housekeeping.rst       |  6 ++-
 2 files changed, 41 insertions(+), 2 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 68647ff4bdd2..eb608e5139a6 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -2822,7 +2822,6 @@ Kernel parameters
 			  "number of CPUs in system - 1".
 
 			managed_irq
-
 			  Isolate from being targeted by managed interrupts
 			  which have an interrupt mask containing isolated
 			  CPUs. The affinity of managed interrupts is
@@ -2845,6 +2844,42 @@ Kernel parameters
 			  housekeeping CPUs has no influence on those
 			  queues.
 
+			managed_irq_strict
+			  Strictly restrict multiqueue hardware queue
+			  allocation and managed interrupt affinity to
+			  housekeeping CPUs. Unlike the best-effort
+			  managed_irq flag, this option guarantees that
+			  managed interrupts and multiqueue block device
+			  queues are never created for or mapped to isolated
+			  CPUs.
+
+			  In the block layer (blk-mq), device queue allocation
+			  and CPU-to-queue mappings are restricted to
+			  housekeeping CPUs. For devices using managed
+			  interrupts, hardware completion interrupts are routed
+			  exclusively to housekeeping cores, shielding isolated
+			  CPUs from I/O interruptions even if they initiated the
+			  request.
+
+			  If both managed_irq and managed_irq_strict are
+			  specified, managed_irq_strict takes precedence.
+
+			  Note: Using managed_irq_strict restricts the number
+			  of allocated hardware queues to match the number of
+			  housekeeping CPUs for single-set drivers. This
+			  prevents MSI-X vector exhaustion and forces isolated
+			  CPUs to share submission queues.
+
+			  Note: Offlining housekeeping CPUs which serve
+			  isolated CPUs will fail. The isolated CPUs must
+			  be offlined before offlining the housekeeping
+			  CPUs.
+
+			  Note: When I/O is submitted by an application on
+			  an isolated CPU, the hardware completion
+			  interrupt is handled entirely by a housekeeping
+			  CPU.
+
 			The format of <cpu-list> is described above.
 
 	iucv=		[HW,NET]
diff --git a/Documentation/core-api/housekeeping.rst b/Documentation/core-api/housekeeping.rst
index ccb0a88b9cb3..6d08c71e6c0e 100644
--- a/Documentation/core-api/housekeeping.rst
+++ b/Documentation/core-api/housekeeping.rst
@@ -23,7 +23,7 @@ CPU. It is simply a CPU that can execute housekeeping work. There must
 always be at least one online housekeeping CPU at any time. The CPUs that
 are not	isolated are automatically assigned as housekeeping.
 
-Housekeeping is currently divided in four features described
+Housekeeping is currently divided in five features described
 by the ``enum hk_type type``:
 
 1.	HK_TYPE_DOMAIN matches the work moved away by scheduler domain
@@ -44,6 +44,10 @@ by the ``enum hk_type type``:
 	to HK_TYPE_DOMAIN except it ignores the isolation performed by
 	cpusets.
 
+5.	HK_TYPE_MANAGED_IRQ_STRICT matches the multiqueue block devices and
+	managed IRQs strictly restricted to housekeeping CPUs through
+	``isolcpus=managed_irq_strict``.
+
 
 Housekeeping cpumasks
 =================================
-- 
2.55.0